SoundID is a powerful automatic content recognition (ACR) platform specially designed for broadcast monitoring.
It enables broadcasters and content providers to log and monitor any number of audio streams (radio, TV,…) from any input and any format. SoundID offers an advanced automatic sound identification solution combined with a robust set of features, e.g. allowing to find recurring sequences in linear broadcasts.
The SoundID solution contains an outstanding report generator fit to create any report by any criteria and filter.
SoundID also provides additional audio recognition services such as discovering new sounds (for automatic database enhancement and marketing support) and focuses on the broadcast confirmation, playlist consolidation, and copyright protection.
How does it work?
Let’s imagine you are a panel member, you have “SoundID Audience” app installed on your smartphone. You don’t have to change any habit, the microphone of your mobile is activated on a regular predefined interval (i.e. once per minute) for a predefined duration (i.e. 5 seconds).
No constraints, the mobile can even stay inside the pocket of your jacket, we can capture enough sound to recognize which radio you are listening to.
Of course, for any personal reasons, you can temporarily disconnect from the panel, knowing that you will lose the reward from the panel owner for this period. In the fact the marginal cost of an extra panel member is very low allowing the panel owner to significantly increase the size of the panel to guarantee that the collected data is representing an effective audience distribution.
Our own fingerprints
Our advantages are numerous: first, no change at all at the source, we simply listen to the live streams to generate the fingerprints by ourselves, then our smartphone app will sample the ambient sounds of the panel user, and there too a fingerprint will be generated locally for each sample (nonreversible process fully protecting the private life) before being uploaded to our central servers and erased locally. Upload fingerprints are then compared to the stream fingerprints to detect which stream the panel user was listening at that time. Full security, low bandwidth, possibly very large panels.
Audience measurement historic
Radio audience measurement is really inaccurate. Until today, the audience is calculated on the basis of the declarative survey. You all know what I’m talking about… live survey over the phone with 25 minutes of questions… A real nightmare both for the surveyor and the panel member. Beyond the fact that it’s annoying, this kind of survey is really not relevant; in the end, the panel member gives any type of answers to speed up the process and terminate the interview as soon as possible.
A second method appeared recently: Audience measurement based on watermarking technology. Other approaches, other problems.
First, the watermarking force a change into the live signal at the source to inject some recurring data to identify the channel listened. This technical step creates an average delay of 150 ms between the live and the broadcasted stream, clearly making the technical team unhappy to have to manage such a situation impacting the chain of the broadcast. Second, if the idea is to ask the panel member to wear a decoder to avoid the manual queries is good, it involves a number of disadvantages: cost of the decoder, an extra device to carry, to recharge … Finally, an expensive solution driving the data collection companies to limit the number of active devices to the minimum. As a real example, 15 devices to cover a region with 230 000 active listeners every day… can we really call that a representative panel?
Our content recognition solution based on home-made complex algorithms is the result of 7 years of research and development and 2 extra years of finetuning to deliver the best result for the dynamic effective audience measurement.
It enables broadcasters and content providers to log and monitor any number of audio streams (radio, TV,…) from any input and any format. SoundID offers an advanced automatic sound identification solution combined with a robust set of features, e.g. allowing to find recurring sequences in linear broadcasts.
The SoundID solution contains an outstanding report generator fit to create any report by any criteria and filter.
SoundID also provides additional audio recognition services such as discovering new sounds (for automatic database enhancement and marketing support) and focuses on the broadcast confirmation, playlist consolidation, and copyright protection.
How does it work?
Let’s imagine you are a panel member, you have “SoundID Audience” app installed on your smartphone. You don’t have to change any habit, the microphone of your mobile is activated on a regular predefined interval (i.e. once per minute) for a predefined duration (i.e. 5 seconds).
No constraints, the mobile can even stay inside the pocket of your jacket, we can capture enough sound to recognize which radio you are listening to.
Of course, for any personal reasons, you can temporarily disconnect from the panel, knowing that you will lose the reward from the panel owner for this period. In the fact the marginal cost of an extra panel member is very low allowing the panel owner to significantly increase the size of the panel to guarantee that the collected data is representing an effective audience distribution.
Our own fingerprints
Our advantages are numerous: first, no change at all at the source, we simply listen to the live streams to generate the fingerprints by ourselves, then our smartphone app will sample the ambient sounds of the panel user, and there too a fingerprint will be generated locally for each sample (nonreversible process fully protecting the private life) before being uploaded to our central servers and erased locally. Upload fingerprints are then compared to the stream fingerprints to detect which stream the panel user was listening at that time. Full security, low bandwidth, possibly very large panels.
Audience measurement historic
Radio audience measurement is really inaccurate. Until today, the audience is calculated on the basis of the declarative survey. You all know what I’m talking about… live survey over the phone with 25 minutes of questions… A real nightmare both for the surveyor and the panel member. Beyond the fact that it’s annoying, this kind of survey is really not relevant; in the end, the panel member gives any type of answers to speed up the process and terminate the interview as soon as possible.
A second method appeared recently: Audience measurement based on watermarking technology. Other approaches, other problems.
First, the watermarking force a change into the live signal at the source to inject some recurring data to identify the channel listened. This technical step creates an average delay of 150 ms between the live and the broadcasted stream, clearly making the technical team unhappy to have to manage such a situation impacting the chain of the broadcast. Second, if the idea is to ask the panel member to wear a decoder to avoid the manual queries is good, it involves a number of disadvantages: cost of the decoder, an extra device to carry, to recharge … Finally, an expensive solution driving the data collection companies to limit the number of active devices to the minimum. As a real example, 15 devices to cover a region with 230 000 active listeners every day… can we really call that a representative panel?
Our content recognition solution based on home-made complex algorithms is the result of 7 years of research and development and 2 extra years of finetuning to deliver the best result for the dynamic effective audience measurement.
Show More
